1. A preliminary study of the Vector Boson Fusion Process with FAMOS
PRS/Higgs meeting 06/08/05
Sara Bolognesi (TORINO)

2. Signal and background definition
PRS/Higgs meeting 06/08/05
Sara Bolognesi (TORINO)

3. Vector Boson Fusion in PHASE
W mn
The signal is
pp qq qqVW qqVW qqqqmn where
V= Z\W qq
PREVIOUS EXISTING MC
PYTHIA
MadEvent
pp qq qqVLWL qqVLWL qqqqmn
pp qq qqVW qqqqmn
only boson fusion diagrams
production and decay approximation (i.e. only diagrams with on shell bosons outgoing)
EVBA approximation
only longitudinal bosons
NEW MC DEVELOPED BY TURIN GROUP
(Accomando, Ballestrero, Maina)
PHASE
pp->qqqqmn O(aEW6)
exact calculation of all the processes
xsec = | fusion diagrams + irreducible background |2
huge interference effects
Reference for PHASE
hep-ph/
C.Mariotti talk (12/08/04) at PRS session of CMS Week
E. Maina talk (05/14/05) at LesHouches Workshop

4. Signal definition: cut
Because of the presence of the irreducible background, we need a definition for signal “a posteriori” (i.e. after events generation) b t W V q q’
Cut against the top contribution*
160 < M(bqq’,bmn) < 190 (GeV)
Ask for two bosons in the final state*
[ 70 < M(mn) < 90 (GeV) ] &&
[ 80 < M(qq) < 100 (GeV) || 70 < M(qq’) < 90 (GeV) ]
Cut against three bosons in the final state*
[ 80 < M(qq) < 100 (GeV) || 70 < M(qq’) < 90 (GeV) ]
for the other 2 quarks
* the right quarks flavours are always requested

5. Signal definition: xsec
M(H) (GeV) no Higgs
total pb pb pb pb
signal pb pb pb pb
top (EW) pb pb pb pb
other irr. backg pb pb pb pb
TOP (EW)
big contribution (at low invariant mass VV)
OTHER IRREDUCIBLE BACKGROUND:
- NON RESONANT
- THREE BOSONS OUTGOING
little contribution (~ constant over all the spectrum)
Higgs mass dependence (because of the Higgsstrahlung diagrams)
2 central q + mn ~ 2 outgoing bosons

6. Signal kinematics p q m n q’ q tag No Higgs sample W V
2 central bosons with high pT
high MET
central muon with high pT
2 q from boson decay: central, little Dh, high pT
2 q tag: high h and big Dh, very high energy

7. Backgrounds
Irreducible background (i.e. background with the same final state of the signal and at the same perturbative order aEW6 )
single and double top (pure EW),
(xsec in a previous slide: ~ 0.5 pb)
non resonant diagrams,
events with three bosons outgoing
2. Background with the same final state of the signal but at a different perturbative order
pp t t 1m+X aS2 aEW pb PYTHIA
pp qqZW qqqqmn still missing (but xsec ~ 3 fb)
pp qqWW qqqqmn aS2 aEW pb MadEvent
pp qqqqW qqqqmn aS4 aEW pb AlpGen
q = u,d,c,s,g
3. Background with lower multiplicity of quarks in the final state (e.g.: Wqq, Wqqq, single W)
not considered to avoid multiple counting due to initial/final state gluon radiation during the parton shower evolution q g g q
Wqq
Wqqqq q g q
We are waiting for the new AlpGen version: it will generate W + n jets without multiple counting W

8. Reco with FAMOS_1_2_0 NO PILE-UP !! GlobalMuons
We try to reconstruct jets with different algorithms
ConeCut =0.5
Iterative Cone Algo
ConeSeedEtCut = 0.5
JetCalibration = GammaJet
JetRecom = 1
MidPointConeRadius= 1
MidPoint Cone Algo
JetInput = EcalPlusHcalTower
MidPointConeSeedThreshold = 0.5
EcalPlusHcalTowerEt = 0.5
KtJet RParameter Algo
RParameter = 1
KtJet DCut Algo
DCut = 400
KtJetAngle = 2
KtJet NJet Algo
NJet = 4
We try to reconstruct MET with different algorithms
FamosMETfromCaloTower
Tower Correction = true
TowerEtCut = 0.5
JetCorrection = true
FamosMETfromJet
JetUnclusterTower = true
JetUnclusterCorrection = true
Primary Vertex Finder
Combined Btagging Algo
Iterative Cone Algo for jets (with the previous parameters)

9. “MC” resolution reconstructed value – generated value resolution (%) =
absolute resolution =
reconstructed value – generated value
(No Higgs sample used as reference in the following)
PRS/Higgs meeting 06/08/05
Sara Bolognesi (TORINO)

10. Resolution: m and n
We choose the closest muon to the generated one (only if DR<0.2)
resolution < 10% in 86% of the events
transverse momentum underestimated
FWHM ~ 1.6 GeV
absolute resolution:
peak at ~ -0.3 GeV
resolution peaked at %
Total generated MET ~ pT of signal n
Backup slide su resolution on pz, resolution MET vs ptgen
METfromJets more precise
resolution < 10% in 19 % of the events
MET underestimated
FWHM ~ 43 GeV
absolute resolution:
peak at ~ -3 GeV
resolution peaked at %
In order to reconstruct pzn we ask for
(pm + pn)2 = mW2

11. Resolution: leptonic W
muon pT and neutrino pT underestimated (previous slide)
W pT underestimated
resolution peaked at %
absolute resolution (gaussian fit):
std. deviation ~ 24 GeV
mean ~ -7 GeV
W pseudorapidity overestimated (absval)
resolution with asymmetric shape

12. Resolution: jets hadronic W/Z
JETS RESOLUTION
Resolution calculated for selected events with only 4 jets: each of them has been matched with the closest generated quark (DR < 1)
Calibration only for ICA !!!
HADRONIC BOSON MASS RESOLUTION
mV resolution peaked at +10% (ICA)
mV absolute resolution (gaussian fit):
std. deviation ~ 18 GeV
mean at ~ 10 GeV

13. Data analysis Work in progress Considered Higgs masses:
300 GeV, 500 GeV, 700 GeV, no Higgs case
PRS/Higgs meeting 06/08/05
Sara Bolognesi (TORINO)

14. Cuts (1) Signal reconstruction signal (NoHiggs) background
(considered only jets with pT > 30 GeV)
muon with maximum pT ( > 20 GeV)
pTn = MET ( > 20 GeV)
leptonic W
74 %
50 %
pzn (pm+pn)2 =mW2
(no cut on mW!)
|hj1|, |hj2| <3
hadronic V
50 < M(j1,j2) <125 (GeV)
52 %
36 %
|hj1-hj2| < 2
(if more then one, the most central has been chosen)
ask for other 2 tag jets = 2 most energetic remaining jets
44 %
25 %

15. Cuts (2) Background rejection signal (NoHiggs) background
(considered only jets with pT > 30 GeV)
cut against top
100 < M(W,q), M(V,q) < 300 (GeV)
38 %
14 %
for any jet with |h| > 3 (out of B-tag algo acceptance)
100 < M(W,b), M(V,b) < 300 (GeV)
b = jet with maximum probability of b-tagging (only if P(b)>1)
pTjtag1,pTjtag2 > 50 GeV
14 %
1.1 %
cuts on tag jets
M(jtag1,jtag2) > 600 GeV
|hjtag1-hjtag2| > 1.5
hjtag1*hjtag2 < -1
|hW| < 2
12 %
0.8 %
some other cuts
M(V,W,jtag1, jtag2) > 1 TeV
(Njet with |hj|<2) < 12
considered also jets with pT < 30 GeV

16. Analysis resolution Comparison between
resolution on the signal after the previous analysis cuts
resolution on the signal using the MC truth
(histos normalized to 1)
With the previous cuts we keep a resolution roughly equal to the “MC” resolution…
… but …

17. Efficiency and significance
… the bad reconstruction resolution affects our efficiency on the signal
remember: we loose 50% of the signal only by asking pTm, pTn > 20 GeV and the presence of an hadronic boson !!!
We would like to be more efficient at high M(VW)
(too many fluctuation: we need more statistic)
Efficiency on the signal and on the backg vs massa
signific vs massa
100 fb-1
100 fb-1
No Higgs fb-1
different backgrounds already summed
The biggest background at high M(VW) after our analysis cuts is Wjjjj

18. Events
Number of events and significance at the Higgs peak ± 30 % ( = our analysis resolution)
100 fb-1
M(H) ± 30% (GeV)
300 ± 90
500 ± 150
700 ± 210
NoH (> 1 TeV)
signal events
1893
1023
473 26
background events
437297
243833
124172
10932
significance
2.9
2.1
1.3
0.2
500 fb-1
M(H) ± 30% (GeV)
300 ± 90
500 ± 150
700 ± 210
NoH (> 1 TeV)
signal events
9463
5113
2364
128
background events
2.186×106
1.219×106
620858
54662
significance
6.4
4.6
3.0
0.5
With this resolution some years are necessary to discover a heavy Higgs boson

19. Problems
Generate W + n jets without double counting with the future AlpGen version
Bad resolution (also when we use the MC truth) first of all for the jets
Not a FAMOS problem (I think)
something similar for top and W mass in ORCA / R.Chierici talk (03/14/05) at PRS Session of CMS Week
Is it a problem related with our reconstruction program (e.g. ConeSeedEtCut too low)?
we have to check, more deeply, different values for the parameter
our hint !
Is it a software problem? (i.e. we have to wait for a more realistic resolution because the software is still in development?)
Or maybe is it a detector problem? Is this the best that we can do with our detector?
We are investigating on…
Plans: to apply the same analysis cuts to the MC truth to “quantify” the resolution problem on the signal efficiency

20. Good news
We have for the first time a MC (PHASE) able to generate the signal in an exact way (without any approximation) and the irreducible background
For a comparison of the different MC see R.Bellan talk (03/15/05) at PRS Session of CMS Week
The background and signal samples we are producing in Turin, are being submitted to the official production
The production status is reported at
An analysis with FAMOS is not only possible but also reliable and VERY fast
In 1 day we have processed events of signal (via LSF)
An ENOURMOUS help from developers always available at
Next-to-next step = to perform the same analysis in ORCA to compare the results

21. Backup slides Generated statistic MV, M(VW) distributions
pzn reconstruction
Detailed efficiency of the cuts
Events and significance integrated over all the spectrum

22. xsec and generated events
signal and irreducible background
M(H) = 300 GeV pb
M(H) = 500 GeV pb
M(H)= 700 GeV pb
no Higgs pb
other backgrounds
pp t t 1m+X pb
pp qqWW qqqqmn pb
pp qqqqW qqqqmn pb
pp qqW+W- qqqqmn pb
pp qqW+W+ qqqqmn pb
pp qqW-W- qqqqmn pb

23. M(V), M(VW) distributions
impossible to distinguish W and Z peaks
100 fb-1
100 fb-1
background histrograms superimposed (not stacked)

24. Neutrino reconstruction
pTn = MET (METFromJets algorithm)
second degree equation with two solutions
pzn calculated with (pm + pn)2 = mW2
we chose the biggest solution and we take the real part if the discriminant is negative
( 30 % of the events )

25. Detailed efficiency of the cuts
cuts on tag jets
some other cuts
leptonic W
hadronic V
2 tag jets
cut top
signal % % % % % %
300 GeV
irred. backg % % % % % %
signal % % % % % %
500 GeV
irred. backg % % % % % %
signal % % % % % %
700 GeV
irred. backg % % % % % %
signal % % % % % %
no Higgs
irred. backg % % % % % %
pp t t 1m+X % % % % % %
pp qqWW qqqqmn % % % % % %
pp qqqqW qqqqmn % % % % % %
For a description of the cuts see slides “Cuts(1)” and “Cuts(2)”

26. Events Events and significance integrated over all the M(VW) spectrum
100 fb-1
M(H) = 300 GeV M(H) = 500 GeV M(H)=700 GeV No Higgs
signal events
2860
2303
2064
1898
background events
771689
771633
771596
771594
significance
3.2
2.6
2.3
2.2
irreducible background depends on the Higgs mass
500 fb-1
M(H) = 300 GeV M(H) = 500 GeV M(H)=700 GeV No Higgs
signal events
14289
11510
10320
9493
background events
~ × 106
significance
7.3
5.9
5.2
4.8
With this resolution some years are necessary to discover a heavy Higgs boson